MIMO DF Relay Beamforming for Secrecy with Artificial Noise, Imperfect CSI, and Finite-Alphabet

In this paper, we consider decode-and-forward (DF) relay beamforming with imperfect channel state information (CSI), cooperative artificial noise (AN) injection, and finitealphabet input in the presence of an user and J non-colluding eavesdroppers. The communication between the source and the user is aided by a multiple-input-multiple-output (MIMO) DF relay. We use the fact that a wiretap code consists of two parts: i) common message (non-secret), and ii) secret message. The source transmits two independent messages: i) common message (non-secret), and ii) secret message. The common message is transmitted at a fixed rate R0, and it is intended for the user. The secret message is also intended for the user but it should be kept secret from the J eavesdroppers. The source and the MIMO DF relay operate under individual power constraints. In order to improve the secrecy rate, the MIMO relay also injects artificial noise. The CSI on all the links are assumed to be imperfect and CSI errors are assumed to be norm bounded. In order to maximize the worst case secrecy rate, we maximize the worst case link information rate to the user subject to: i) the individual power constraints on the source and the MIMO relay, and ii) the best case link information rates to J eavesdroppers be less than or equal to R0 in order to support a fixed common message rate R0. Numerical results showing the effect of perfect/imperfect CSI, presence/absence of AN with finite-alphabet input on the secrecy rate are presented. keywords: MIMO relay beamforming, physical layer security, multiple eavesdroppers, artificial noise, imperfect CSI, finite-alphabet input, semidefinite programming.